Can phonemic mergers reverse?

[Space60]

Can a region that historically had a merger now make a distinction due to the merger reversing? Does it depending on how widespread the merger is? Occurring in a small community vs. millions of speakers?

[WeepingElf]

Mergers can't reverse, but of course, a phoneme resulting from a merger may split again, even such that the original inventory is restored, but the distribution of the new phonemes will be different from that of the old ones.

[Linguoboy]

A distinction lost due to a phonemic merger could be reintroduced in a variety either on the basis of a written standard which preserves the status quo ante or of a related variety which lacked the merger.

I think in either case, it's not going to be a perfect reversal (since the written standard never perfectly preserves the spoken language and even very closely related varieties differ in unpredictable ways), but it could be close enough as makes no nevermind.

[Kuchigakatai]

Can a region that historically had a merger now make a distinction due to the merger reversing? Does it depending on how widespread the merger is? Occurring in a small community vs. millions of speakers?

Much of Andalusia now distinguishes /s/ from /θ/ under the pressure of other dialects in the Peninsula.

For example, here is "Adelita Power", a 25 years old girl from Andalusia (22 in the video), distinguishing /s/ and /θ/ while maintaining other dialectal features such as great amounts of debuccalization and some dropping of both of those phonemes: https://www.youtube.com/watch?v=PJd51H_uF8k

Partial transcription:
(Beginning at 1:07)
La gente se cree que por tú ser andaluza, tienes que ser maja,
[la ˈxente se ˈkɾe.e ke poɾ ˈtu seɾ andaˈluθa ˈtjene ke seɾ ˈmaha]
'People [elsewhere in Spain] think that because you're Andalusian, you have to be charming,'

tienes que ser simpática todo el rato, graciosa.
[ˈtjene ke seɾ simˈpatika toðo eɾ ˈrato ɣɾaˈθjosa]
'you have to be likeable all the time, funny.'

Y las cosas no son así, ¿eh?
[i lah ˈkosah no soŋ aˈsi ˈe]
'But things are not like that, okay!?'

¡Los andaluces también tenemos sentimientos!
[loh andaˈluθe tamˈbjen teˈnemo sentiˈmjentoh]
'We Andalusians also have feelings!'

¡Los andaluces también tenemos días buenos y días malos!
[loh andaˈluθeh tamˈbjen teˈnemoh ˈðjah ˈβwoh i ˈðjah ˈmaloh]
'We Andalusians also have good days and bad days!'

¡Los andaluces...! ...De verdad que somos muy majos, y muy graciosos la mayoría.
[loh andaˈluθeh ... de βeɾˈða ke ˈsomoh muj ˈmaxoh i muj ɣɾaˈθjoso la majoˈɾi.a]
'We Andalusians...! ...Honestly we're mostly very charming and very funny.'

Pero bueno, hay de todo.
[peɾo ˈβweno ˈai ðe ˈto]
'But anyway, there's bit of everything.'

No estamos cortados todos por la misma tijera.
[no ˈtʰamo koɾˈtao ˈtʰo pʰoɾ la mihma tiˈheɾa]
'We're not all cut out with the same scissors.'

(Menos los de Granada tienen más mala folla, eso está claro.)
[men loh ðe ɣɾaˈnaða ˈtjeneŋ ˈmah ˈmala ˈfoja, eθo ehta ˈklaɾo]
'(Except for those from Granada, they are more evil*, that is clear.)'

*Literally "they have worse sex", but this expression is very idiomatic and pretty much never has its literal meaning.

I think she actually pronounces eso as [eθo] near the end of this fragment, and in the next bit she also seems to say siesta with [θ] for the first /s/, so for some items the reappearance of the distinction may not be perfect.

[WeepingElf]

Oh, yes, I forgot the possibility that a distinction lost in one dialect can be restored under the influence of another dialect that never lost it. Insofar my statement that "mergers can't be reversed" needs to be qualified. Also, it may be the case that the old distinction has left traces elsewhere (e.g., in preceding or following segments) which spread back to the merged phoneme, causing it to split again along the old distinction.

[zompist]

Also, FWIW, it's possible to have a near-merger that's reversed. That is, you might have two sounds that (say) people from another variety can't tell apart, but which the speakers themselves somehow do.

I remember this from Labov, and his example had to do with lax and tense vowels in English, but I don't recall the details.

[Salmoneus]

This very often happens with regional dialects, losing distinctive features due to relatively low prestige of the dialects that those features too-uniquely identify.

Off-hand, two obvious examples in English would be the (almost complete?) loss of the curl-coil merger in New York - although that's at least in part because the New York dialects have have to perpetuate themselves through continual recruitment of new speakers migrating from other dialects, which must have made it harder - and the ongoing loss of the thick-hick merger in certain urban Scottish dialects. In the former case, the merger became stigmatised by outsiders, which lead to its reversal due to shame. [an earlier parallel in the UK might be the restoration (and widespread hyperrestoration) of initial /h/ in many English dialects]. This is relatively straightforward - New Yorkers simply had to re-learn the distinction by observing outsiders, producing lots of hypercorrections along the way.

The Scottish case may be more interesting, because the /T/-/h/ merger is not just being undone, but being rapidly replaced by the /T/-/f/ merger - and my ill-informed understanding is that it isn't accompanied by the same sort of hypercorrections. The easiest explanation may be that although speakers merged the pronunciations, they still understood that there were two lexical sets, and were simply choosing to merge them; now they're just choosing to merge one set with a different set instead. This is probably helped out by a) the fact that most other dialects don't have the /T/-/h/ merger, and b) the fact that the distinction is very clear in writing.


Something weird has also perhaps happened in SSBE. SSBE is usually considered the descendent of RP, and most of the differences can be explained in that fashion. However, RP had tensing of /Q/, the creation of a new set, and the subsequent merger of that set with /O/. SSBE has seemingly reversed the cloth-thought merger, and then merged cloth back into lot. Even though RP was the prestige dialect while this was happening, and the most prestigious non-local dialects (those of America) agree with RP in this. So did RP spontaneously reverse this sound change - without, so far as I'm aware, any hypercorrection - or is SSBE not actually descended from RP at all, but from a very similar and unrecorded 'vulgar English', as it were, that never underwent these shifts?

[anteallach]

Something weird has also perhaps happened in SSBE. SSBE is usually considered the descendent of RP, and most of the differences can be explained in that fashion. However, RP had tensing of /Q/, the creation of a new set, and the subsequent merger of that set with /O/. SSBE has seemingly reversed the cloth-thought merger, and then merged cloth back into lot. Even though RP was the prestige dialect while this was happening, and the most prestigious non-local dialects (those of America) agree with RP in this. So did RP spontaneously reverse this sound change - without, so far as I'm aware, any hypercorrection - or is SSBE not actually descended from RP at all, but from a very similar and unrecorded 'vulgar English', as it were, that never underwent these shifts?

This is a bit different from the others as both types have the same two phonemes but CLOTH words go with LOT in one and THOUGHT in the other. I think it's easier for words to be transferred from one phonological category to another than for a phonological category to be split back into two again. However, I don't know why several <aus> words, e.g. Austria, Australia, sausage, have LOT/CLOTH rather than the THOUGHT you might expect from the spelling; are they actually hypercorrections?

My impression is that for speakers born in (approximately) the first half of the 20th century the LOT-CLOTH split was a "U-RP" thing, as in something that you'd find in out and out aristocratic speech but not so much in general RP. (Of course the Queen is known for it.) It is also, IIRC, a feature of the traditional accents of southern England covering much the same area as the TRAP-BATH split, though it has also been lost from most non-RP/SSBE speech in southern England too. (And it's also either absent or been largely lost in Australian and New Zealand English.) To better understand what happened, and why the LOT-CLOTH split disappeared whereas the TRAP-BATH one survived, I think you'd need a better idea of the sociolinguistics of the split in England in the late 19th or early 20th century than I have.

[Space60]

Something weird has also perhaps happened in SSBE. SSBE is usually considered the descendent of RP, and most of the differences can be explained in that fashion. However, RP had tensing of /Q/, the creation of a new set, and the subsequent merger of that set with /O/. SSBE has seemingly reversed the cloth-thought merger, and then merged cloth back into lot. Even though RP was the prestige dialect while this was happening, and the most prestigious non-local dialects (those of America) agree with RP in this. So did RP spontaneously reverse this sound change - without, so far as I'm aware, any hypercorrection - or is SSBE not actually descended from RP at all, but from a very similar and unrecorded 'vulgar English', as it were, that never underwent these shifts?

This is a bit different from the others as both types have the same two phonemes but CLOTH words go with LOT in one and THOUGHT in the other. I think it's easier for words to be transferred from one phonological category to another than for a phonological category to be split back into two again. However, I don't know why several <aus> words, e.g. Austria, Australia, sausage, have LOT/CLOTH rather than the THOUGHT you might expect from the spelling; are they actually hypercorrections?

I think those were hypercorrections that occurred when the split disappeared in the British dialects that had it. I've also read about hypocorrections occurring. Some traditional Australian English dialects that had mostly lost the lot-cloth split still had the THOUGHT vowel in the words "off" and "gone" from what I've read.

[Hominid]

"Curl" and "coil" were never actually merged though, were they? I thought the vowel in "curl" in that dialect was actually more like the one in "light" with Canadian raising.

[Ryusenshi]

So did RP spontaneously reverse this sound change - without, so far as I'm aware, any hypercorrection - or is SSBE not actually descended from RP at all, but from a very similar and unrecorded 'vulgar English', as it were, that never underwent these shifts?

I think the CLOTH-THOUGHT merger only affected the upper classes, while middle-class London accents continued to have CLOTH=LOT: as anteallach said, CLOTH=THOUGHT was more of a "U-RP" thing. Besides, RP and middle-class London (or "Estuary") have never been completely separated from one another. Isn't SSBE more or less a mixture of the two? AIUI, several innovations of SSBE, like the wholly/holy split or the poor/pour merger, originated in Estuary English then moved "upwards". A conservative trait can certainly do the same, especially if the alternative gets stigmatized as "too posh".

[Travis B.]

So did RP spontaneously reverse this sound change - without, so far as I'm aware, any hypercorrection - or is SSBE not actually descended from RP at all, but from a very similar and unrecorded 'vulgar English', as it were, that never underwent these shifts?

I think the CLOTH-THOUGHT merger only affected the upper classes, while middle-class London accents continued to have CLOTH=LOT: as anteallach said, CLOTH=THOUGHT was more of a "U-RP" thing. Besides, RP and middle-class London (or "Estuary") have never been completely separated from one another. Isn't SSBE more or less a mixture of the two? AIUI, several innovations of SSBE, like the wholly/holy split or the poor/pour merger, originated in Estuary English then moved "upwards". A conservative trait can certainly do the same, especially if the alternative gets stigmatized as "too posh".

The question I have, then, is why did CLOTH=THOUGHT become completely dominant in North America then if it was at most affected in England?

[Space60]

The lot-cloth split is universal among North American varieties that lack that cot-caught merger as far as I know. Also North American English extended the lot-cloth split to positions before velars in "long", "dog" and "chocolate" which never occurred in the British version of the split. The split must have occurred centuries ago in Britain, elsewise North Americans likely wouldn't have it.

[vegfarandi]

A famous instance is the reversal of flámæli in Icelandic, which involved the merger of /ɪ/ + /ɛ/ to /e/ and /ʏ/ + /œ/ to /ø/. This was nearly completely wiped out by the middle of the 20th century using the iron fist of centralized education curricula.

[Nortaneous]

The lot-cloth split is universal among North American varieties that lack that cot-caught merger as far as I know. Also North American English extended the lot-cloth split to positions before velars in "long", "dog" and "chocolate" which never occurred in the British version of the split. The split must have occurred centuries ago in Britain, elsewise North Americans likely wouldn't have it.

I don't think I've ever heard THOUGHT in "chocolate", except from New Englanders with the LOT-THOUGHT merger.

[zompist]

I don't think I've ever heard THOUGHT in "chocolate", except from New Englanders with the LOT-THOUGHT merger.

Really? ch[ɔ]colate is normal for me... haven't you noticed the folksy spelling "chawklit"?

(I wonder if there's a linguistic term for that, BTW... faux-dialectal spellings like "sez".)

[akam chinjir]

(I wonder if there's a linguistic term for that, BTW... faux-dialectal spellings like "sez".)

Pronunciation spellings? (Kidding.)

Another example is the use of "-in'" in place of "-ing" for the verbal ending, encoding more the class or dialect or whatever of the speaker than anything about the pronunciation (given that [n] here has a far wider distribution than is ever reflected in this spelling).

[anteallach]

A famous instance is the reversal of flámæli in Icelandic, which involved the merger of /ɪ/ + /ɛ/ to /e/ and /ʏ/ + /œ/ to /ø/. This was nearly completely wiped out by the middle of the 20th century using the iron fist of centralized education curricula.

What was its social/regional distribution?

[anteallach]

So did RP spontaneously reverse this sound change - without, so far as I'm aware, any hypercorrection - or is SSBE not actually descended from RP at all, but from a very similar and unrecorded 'vulgar English', as it were, that never underwent these shifts?

I think the CLOTH-THOUGHT merger only affected the upper classes, while middle-class London accents continued to have CLOTH=LOT: as anteallach said, CLOTH=THOUGHT was more of a "U-RP" thing. Besides, RP and middle-class London (or "Estuary") have never been completely separated from one another. Isn't SSBE more or less a mixture of the two? AIUI, several innovations of SSBE, like the wholly/holy split or the poor/pour merger, originated in Estuary English then moved "upwards". A conservative trait can certainly do the same, especially if the alternative gets stigmatized as "too posh".

The question I have, then, is why did CLOTH=THOUGHT become completely dominant in North America then if it was at most affected in England?

It wasn't; it was also a feature of the traditional dialects of most of southern England, which will be where AmE got it from. The accents which didn't have it were more mainstream RP (as opposed to "U-RP") and near-RP (and also anything in the North). I don't know about long, dog, etc.

[Space60]

I think the CLOTH-THOUGHT merger only affected the upper classes, while middle-class London accents continued to have CLOTH=LOT: as anteallach said, CLOTH=THOUGHT was more of a "U-RP" thing. Besides, RP and middle-class London (or "Estuary") have never been completely separated from one another. Isn't SSBE more or less a mixture of the two? AIUI, several innovations of SSBE, like the wholly/holy split or the poor/pour merger, originated in Estuary English then moved "upwards". A conservative trait can certainly do the same, especially if the alternative gets stigmatized as "too posh".

The question I have, then, is why did CLOTH=THOUGHT become completely dominant in North America then if it was at most affected in England?

It wasn't; it was also a feature of the traditional dialects of most of southern England, which will be where AmE got it from. The accents which didn't have it were more mainstream RP (as opposed to "U-RP") and near-RP (and also anything in the North). I don't know about long, dog, etc.

As far as I know, "long" and "dog" never had the THOUGHT vowel in any British dialects. This was a North American innovation. Speaking of "-og" words they tend to vary in Americans that have the lot-cloth split. "Dog" almost always has the THOUGHT vowel for those with the split other words like "log", "hog", "frog", "blog" etc. are variable.